Methods and Technologies for Pedestrian and Bicycle Volume Data Collection NCHRP 7-19

NATMEC: Bike and Pedestrian Detection

July 1, 2014

Presentation Overview

Introduction

Guidebook Walkthrough

Testing Approach and Findings

Final Remarks

2

Project Purpose

Address lack of pedestrian and bicycle volume data

–Barrier to planning effective facilities

–Standard procedures for vehicular data collection

Assess variety of existing and new technologies and methods

Develop guidance for practitioners

3

Guidebook Purpose

Guidebook produced as a resource for practitioners

Designed to help practitioners:

–Understand the value of multimodal data

–Develop a data collection plan

–Identify and recommend data collection methods

–Correct raw count data from a particular technology

4

Guidebook Organization

Quick Start Guide

1. Introduction

2. Non-Motorized Count Data Applications

3. Data Collection Planning and Implementation

4. Adjusting Count Data

5. Sensor Technology Toolbox Case Studies

Manual Pedestrian and Bicyclist Counts: Example Data Collector

Instructions

Count Protocol Used for NCHRP Project 07-19

Appendix D. Day-of-Year Factoring Approach

5

Ap

pen

dic

es

2. Non-Motorized Count Applications

6

Measuring facility usage

Evaluating before-and-after data

Monitoring travel patterns

Safety analysis

Project prioritization

Multimodal modeling

Source: Kittelson & Associates, Portland State University, and

Toole Design Group (2012)

Before-and-After Bicycle Facility Usage – buffered bicycle lanes on Pennsylvania Avenue

For each application:

Details Case Studies

3. Data Collection Planning & Implementation

7

Covers:

1. Planning the count program

2. Implementing the count program

Provides examples, detailed guidance, checklists

Source: Tony Hull, Toole Design Group.

4. Adjusting Count Data

8

Sources of counter inaccuracy

Measured counter accuracy

Counter correction factors

Expansion factors

Examples applications

Occlusion error

5. Treatment Toolbox

9

Description

Typical application

Level of effort

Strengths

Limitations

Accuracy

Usage

Sidebar with quick facts

Testing Plan

Focus on testing and evaluating commercially available automated technologies

Assess type of technology as opposed to a specific product

Cover a range of facility types, mix of traffic, and geographic locations

Evaluate accuracy through the use of manual count video data reduction

10

Technologies and Site Locations

Technologies

–Passive infrared

–Active infrared

–Pneumatic tubes

–Inductive loops

–Piezoelectric

–Radio beam

11

Site Locations – Portland, OR

– San Francisco, CA

– Davis, CA

– Berkeley, CA

– Minneapolis, MN

– Washington, D.C.

– Arlington, VA

– Montreal, Canada

Video Data Collection

Camera installed with counters for ~5 days

Second deployment targeting desired conditions

~3k hours of video collected

Source: Karla Kingsley, Kittelson & Associates, Inc.

Example site: Portland, OR

Eastbank Esplanade

Multiuse path

Tested:

–Passive Infrared

–Pneumatic Tubes

–Radio Beam

13

Graphical Analysis

14

Undercounting

Overcounting

𝐴𝑃𝐷 = 1

𝑛

𝐴𝑡−𝑀𝑡

𝑀𝑡

𝑛𝑡=1

𝐴𝐴𝑃𝐷 = 1

𝑛

𝐴𝑡−𝑀𝑡

𝑀𝑡

𝑛𝑡=1

𝑟 = (𝑀𝑡−𝑀 )(𝐴𝑡−𝐴 )𝑛𝑡=1

𝑀𝑡−𝑀 2𝑛

𝑡=1 𝐴𝑡−𝐴 2𝑛

𝑡=1

Where Ai is the automated count in period i and Mi is the manual count in period i

Accuracy Calculations

Passive Infrared (IR)

Detect pedestrians and cyclists by infrared radiation (heat) patterns them emit

Passive infrared sensor placed on one side of facility

Widely used and tested

16

Source: Ciara Schlichting, Toole Design Group

Passive Infrared

Easy installation

Mounts to existing pole/surface or in purpose-built pole

Potential false detections from background

Possible undercounting due to occlusion

17

Photo: Frank Proulx

Passive Infrared Findings

18

APD = -8.75%, AAPD = 20.11%, r = 0.9502

Differences between products

Correction function could account for facility width

Accuracy not affected by high temperatures

Active Infrared (IR)

Transmitter and receiver with IR beam

Counts caused by “breaking the beam”

Moderately easy installation – requires aligning transmitter and receiver

19

Source: Steve Hankey, University of Minnesota

Active Infrared

APD = -9.11%

AAPD = 11.61%

r = 0.9991

Single device tested – accurate and highly precise

20

Pneumatic Tubes

One or more tubes are stretched across roadway or path

When a bicycle rides over tube, pulse of air passes through tube to detector

21

Source: Karla Kingsley, Kittelson & Associates, Inc.

Pneumatic Tubes Findings

APD = -17.89%, AAPD = 18.50%, r = 0.9864

Strong site and device specific effects

Accuracy rates not observed to decline with aging tubes

Future research in mixed traffic settings

22

Inductive Loops

Generate a magnetic field that detect metal parts of bicycle passing over loop

In-pavement or temporary loops (on surface)

23

Source: Katie Mencarini, Toole Design Group

Inductive Loops

Permanent (in ground) or temporary (on surface)

Bypass errors

–Cyclists passing

outside bike lane

–Loops leaving gaps

in detection zone

24

Inductive Loops

APD = 0.55%, AAPD = 8.87%, r = 0.9938

Errors with age of loops not detected

Higher volumes slightly affect accuracy

No substantial difference between permanent and temporary loops

25

Inductive Loops

Need to mitigate bypass errors

26

Piezoelectric Sensor

Emit an electric signal when physically deformed to detect bicyclists

Typically embedded in pavement across travel way

27

Source: MetroCount

Piezoelectric Strips

Tested one existing device, due to difficulties procuring equipment

CAUTION – data from single device not installed by research team

APD = -11.36%, AAPD = 26.60%, r = 0.691

28

Radio Beam

Transmitter and receiver emit a radio signal that detect a user when the beam is broken

Not previously tested in literature

Some devices count bikes and peds separately

29

Source: Karla Kingsley, Kittelson & Associates, Inc.

Radio Beam

Product B higher accuracy

Product A – low precision and lower accuracy

Occlusion errors

30

Recommendations for Practitioners

Calibrate and conduct your own ground-truth count tests

Consider approvals and site characteristics when selecting a count site

31

Suggested Research

Additional testing of automated technologies

–Technologies not tested or underrepresented

–Additional sites and conditions

Extrapolating short-duration counts to longer-duration counts

Adjustment factors for environmental factors

32

Questions?

Contact Information

–Kelly Laustsen / klaustsen@kittelson.com / 503.535.7439

–Frank Proulx / fproulx@berkeley.edu

33